C12.ai Closes ¥50M Round as Robots Effectively Enter the Laboratory
C12.ai completes a new ¥50M funding round led by Yonghua Capital. Its flagship product Talos has been running continuously for 50 days inside BeOne Medicines' lab, performing over 3,000 physical operations with zero execution errors — demonstrating proficiency on par with senior medicinal chemists.
After a Chinese New Year saturated with embodied intelligence, new funding and product announcements are arriving at a breakneck pace. While humanoid robots folding laundry, tightening screws, delivering packages, and doing backflips go viral, a quieter experiment has been running for 50 days inside a BeOne Medicines laboratory.
Talos, a mobile-robot-form "Physical AI Scientist," has been performing molecular purification in BeOne Medicines' lab — a tedious yet time-consuming foundational task that previously consumed over 50% of human medicinal chemists' working hours and demands significant experience and manual dexterity.
During the 50-day trial, Talos demonstrated operational proficiency approaching that of a senior medicinal chemist — an industry first. Across the 15 molecular purification experiments completed so far, each molecule involved over 200 consecutive operation steps — totaling more than 3,000 physical operations with zero execution errors.
Waves has exclusively learned that the developer behind Talos, C12.ai, has just completed a new ¥50M funding round led by Yonghua Capital, with participation from Weihao Chuangxin and individual investors. C12.ai's initial round was completed in August 2022, led by Yunqi Partners, with co-investment from Jingya Capital, BioMap, and individual investors.
Standardized Hardware, Software-Defined Capabilities
The robot form factor is not what makes Talos noteworthy — its hardware is a standard dual-arm mobile robot available for lease. The key lies in the brain C12.ai has built for it: through proprietary sensors, the company has trained professional experiment operation skills and drug development prediction capabilities on a general-purpose robot.
Talos not only autonomously executes end-to-end, cross-instrument experiment workflows like a skilled technician, but can also perform real-time reasoning, planning, and error recovery during experiments through its proprietary Vision-Language-Action (VLA) architecture. In specific scenarios, its professional capabilities have reached the level of a human expert with 10 years of experience, handling the bulk of repetitive foundational lab work and freeing scientists to focus on more creative research decisions.
Why Start with Molecular Purification?
Approximately 50% of drug development time is spent on purification. Previously, after initiating a reaction, scientists might not receive the product until the third day. Now with the robot, purification can be completed the same evening, with the next step beginning the following day — cutting the R&D cycle in half.
Unlike traditional automation's rigid "one machine, one task" model, Talos is a goal-driven general-purpose robot that operates existing instruments in existing labs. It moves between workstations like a human scientist, with real-time adaptation and error recovery capabilities.
C12.ai's capabilities for the robot operate on two levels: first, operational skills — using vision and force perception to control robotic arms for delicate movements; second, professional judgment — such as predicting molecular polarity, supported by specialized models. Together, these form a self-iterating closed loop.
VLA Architecture: Not Just Execution, but Reasoning
C12.ai equipped Talos with a VLA architecture that enables the robot to go beyond executing preset programs to real-time reasoning and planning. Given an experiment intent, it first identifies the intent, manages permissions, and then autonomously generates a detailed experiment plan. On the execution side, it perceives the environment in real time and adapts — if deviations occur during an experiment, it can self-recover from errors.
On the safety front, lab tolerance for error is extremely low — even stricter than home environments. Talos possesses advanced environmental awareness: when a human scientist approaches, it automatically stops to ensure absolute human-robot collaboration safety.
On the business execution front, 14 of 15 molecules were purified successfully — a 93% success rate. The single exception was due to the molecule's solubility exceeding the current method's applicable range, not a robot operational error. The current focus is handling 80% of routine molecular purification; for the remaining 20% of highly complex cases, the system identifies and flags them in real time rather than forcing trial-and-error.
From Purification to Full Pipeline: Expanding Along the Drug Lifecycle
Purification is merely the first entry point into the laboratory. Founder and CEO Chen Zhigang stated that the company's roadmap follows the drug development lifecycle: currently in the early R&D stage — retrosynthetic AI plus small-molecule purification; next, moving into preclinical ADME (pharmacokinetic) experiments and process development; then GxP-compliant manufacturing and analytical method automation in the clinical stage; and ultimately, HMLV flexible production lines with online QC and continuous process verification in the commercialization stage.
Beyond pharmaceuticals, C12.ai is exploring flexible cosmetics manufacturing. A consumer takes a photo, AI identifies skin conditions and generates a custom formula, and the robot produces a small batch overnight for direct delivery. "Any physical scenario that requires small-batch, multi-lot continuous manufacturing with environmental adaptability is our target," said Chen.
A Cross-Industry Veteran's Entrepreneurial Logic
Founder and CEO Chen Zhigang is a veteran spanning internet healthcare and pharmaceutical manufacturing. He previously served as Chief Architect / Head of Algorithm and Modeling at Alibaba Health, founded Tencent's Medical Big Data Lab, and was WuXi AppTec's first Chief Digital Officer (CDO).
During his tenure as CDO at WuXi AppTec, he led a team that used algorithms to raise production equipment utilization from 40% to 60%. However, traditional automation equipment — purpose-built workstations — had fatal weaknesses: rigidity, high cost, and inability to be reused across scenarios.
"Traditional automation solved the standardization problem, but flexibility has always been its weakness," said Chen. "What we're building is a general-purpose, software-defined robot scientist."
With this funding round complete, C12.ai will accelerate product iteration and multi-scenario deployment. "We're not building a single-point tool — we're letting the robot penetrate step by step along the entire life of a drug. Ultimately, we want better medicines to reach patients faster and more affordably."